Appendix A: Data Sources and Schema

Data Sources

Source Description Format Size
Our World In Data Energy Generation/Consuption CSV 22.79MB
Our World In Data CO2 Emissions CSV 31.80MB

Data Schema

CO2 Dataset:

Column_Name Data Type
country character
year numeric
co2 numeric

Energy Dataset:

Column_Name Data Type
country character
year numeric
gdp numeric
electricity_generation numeric

Appendix B: Data Cleaning and Transformation Steps

Handling Missing Values:

filter_data <- function(df) {
  df %>%
    filter(!str_starts(iso_code, "OWID"), !is.na(country), !is.na(year)) %>%
    filter(year >= 2000) %>%
    select(where(~ sum(is.na(.)) < 0.4 * nrow(df))) %>%
    filter(!is.na(population), !is.na(gdp))
}
  • Removing columns with 60% missing values.

Selecting only useful columns:

Energy Dataset:

select_energy_columns <- function(df) {
  df %>%
    select(
      country, iso_code, year, population, gdp,
      electricity_generation,
      renewables_electricity, fossil_electricity,
      solar_electricity, wind_electricity, hydro_electricity,
      renewables_share_elec, coal_share_elec, gas_share_elec, oil_share_elec
    )
}

CO2 Dataset:

co2_data = co2_data %>% select(-c(cement_co2,cement_co2_per_capita,co2_growth_abs,co2_growth_prct,co2_including_luc_growth_abs,co2_including_luc_growth_prct,cumulative_cement_co2,cumulative_co2_including_luc,cumulative_luc_co2,flaring_co2,flaring_co2_per_capita,cumulative_flaring_co2,share_global_flaring_co2,share_global_cumulative_flaring_co2))

Transforming the data and adding new columns:

CO2 Dataset:

#CO₂ emissions year-over-year change
co2_data = co2_data %>%
  group_by(country) %>%
  mutate(co2_pct_change = (co2 - lag(co2)) / lag(co2) * 100)
#Total CO₂ from fossil fuel types 
co2_data = co2_data %>%
  mutate(fossil_fuel_co2 = coal_co2 + oil_co2 + gas_co2)
#log gdp
co2_data = co2_data %>% mutate(log_gdp = log(gdp))
#log co2
co2_data = co2_data %>% mutate(log_co2 = log(co2))

Energy Dataset:

add_normalized_metrics <- function(df) {
  df %>%
    mutate(
      electricity_per_capita = electricity_generation / population,
      renewables_per_capita = renewables_electricity / population,
      fossil_per_capita = fossil_electricity / population,
      electricity_per_gdp = electricity_generation / gdp,
      gdp_per_electricity = gdp / (electricity_generation + 1)
    )
}

add_log_transforms <- function(df) {
  df %>%
    mutate(
      log_gdp = log(gdp + 1),
      log_population = log(population + 1),
      log_electricity = log(electricity_generation + 1)
    )
}

na_percentage <- function(df) {
  sapply(df, function(col) round(mean(is.na(col)) * 100, 2))
}

add_energy_ratios <- function(df) {
  df %>%
    mutate(
      fossil_to_renewable_ratio = fossil_electricity / (renewables_electricity + 1),
      fossil_share_elec = fossil_electricity / (electricity_generation + 1),
      solar_share = solar_electricity / (renewables_electricity + 1),
      wind_share = wind_electricity / (renewables_electricity + 1),
      hydro_share = hydro_electricity / (renewables_electricity + 1)
    )
}

add_classification_flags <- function(df) {
  df %>%
    mutate(
      high_renewable = if_else(renewables_share_elec > 50, 1, 0),
      transitioning = if_else(renewables_share_elec > fossil_share_elec, 1, 0)
    )
}

  • Applied log transformations

  • Normalized some columns

  • Added ratios of different columns

Appendix C: Deduplication

energy_transformed %>% filter(duplicated(.))

Appendix D: Code for Reproducibility

Retrieving, Cleaning and Merging the Data sets:

#import these libraries
library(tidyverse)
library(skimr)
library(janitor)
library(naniar)
library(ggplot2)
library(corrplot)
library(broom)
library(countrycode)

#Functions for loading the energy Data
load_energy_data <- function(url = "https://raw.githubusercontent.com/owid/energy-data/master/owid-energy-data.csv") {
  read_csv(url)
}

#function for removing columns with too many missing values
filter_energy_data <- function(df) {
  df %>%
    filter(!str_starts(iso_code, "OWID"), !is.na(country), !is.na(year)) %>%
    filter(year >= 2000) %>%
    select(where(~ sum(is.na(.)) < 0.4 * nrow(df))) %>%
    filter(!is.na(population), !is.na(gdp))
}

#selecting relevant columns
select_energy_columns <- function(df) {
  df %>%
    select(
      country, iso_code, year, population, gdp,
      electricity_generation,
      renewables_electricity, fossil_electricity,
      solar_electricity, wind_electricity, hydro_electricity,
      renewables_share_elec, coal_share_elec, gas_share_elec, oil_share_elec
    )
}

#normalizing columns on population and 1 on gdp
add_normalized_metrics <- function(df) {
  df %>%
    mutate(
      electricity_per_capita = electricity_generation / population,
      renewables_per_capita = renewables_electricity / population,
      fossil_per_capita = fossil_electricity / population,
      electricity_per_gdp = electricity_generation / gdp,
      gdp_per_electricity = gdp / (electricity_generation + 1)
    )
}

#log transfoming columns
add_log_transforms <- function(df) {
  df %>%
    mutate(
      log_gdp = log(gdp + 1),
      log_population = log(population + 1),
      log_electricity = log(electricity_generation + 1)
    )
}

#creating ratio columns
add_energy_ratios <- function(df) {
  df %>%
    mutate(
      fossil_to_renewable_ratio = fossil_electricity / (renewables_electricity + 1),
      fossil_share_elec = fossil_electricity / (electricity_generation + 1),
      solar_share = solar_electricity / (renewables_electricity + 1),
      wind_share = wind_electricity / (renewables_electricity + 1),
      hydro_share = hydro_electricity / (renewables_electricity + 1)
    )
}

#creating catagorical columns
add_classification_flags <- function(df) {
  df %>%
    mutate(
      high_renewable = if_else(renewables_share_elec > 50, 1, 0),
      transitioning = if_else(renewables_share_elec > fossil_share_elec, 1, 0)
    )
}

#function to run previous transformation functions
transform_energy_data <- function(df) {
  df %>%
    add_normalized_metrics() %>%
    add_log_transforms() %>%
    add_energy_ratios() %>%
    add_classification_flags()
}


energy_raw <- load_energy_data()
energy_clean <- filter_energy_data(energy_raw)
energy_selected <- select_energy_columns(energy_clean)
energy_transformed <- transform_energy_data(energy_selected)

#create continent column
energy_transformed <- energy_transformed %>%
  mutate(continent = countrycode(country, "country.name", "continent"))

#loading CO2 data
co2_data <- read_csv("https://raw.githubusercontent.com/owid/co2-data/master/owid-co2-data.csv")

#filtering CO2 data
filter_co2_data <- function(df) {
  df %>%
    filter(!str_starts(iso_code, "OWID"), !is.na(country), !is.na(year)) %>%
    filter(year >= 2000) %>%
    select(where(~ sum(is.na(.)) < 0.4 * nrow(df))) %>%
    filter(!is.na(population), !is.na(gdp))
}

co2_data = filter_co2_data(co2_data)

#selecting relevant columns
co2_data = co2_data %>% select(-c(cement_co2,cement_co2_per_capita,co2_growth_abs,co2_growth_prct,co2_including_luc_growth_abs,co2_including_luc_growth_prct,cumulative_cement_co2,cumulative_co2_including_luc,cumulative_luc_co2,flaring_co2,flaring_co2_per_capita,cumulative_flaring_co2,share_global_flaring_co2,share_global_cumulative_flaring_co2))


##Creating new columns and transforming columns

#CO₂ emissions year-over-year change
co2_data = co2_data %>%
  group_by(country) %>%
  mutate(co2_pct_change = (co2 - lag(co2)) / lag(co2) * 100)
#Total CO₂ from fossil fuel types 
co2_data = co2_data %>%
  mutate(fossil_fuel_co2 = coal_co2 + oil_co2 + gas_co2)
#log gdp
co2_data = co2_data %>% mutate(log_gdp = log(gdp))
#log co2
co2_data = co2_data %>% mutate(log_co2 = log(co2))

Appendix E: Data Cleaning and Processing Summary

Data Cleaning Impact Overview:

CO2 Dataset:

Data Wrangling Step Rows Columns Missing Values
Raw Dataset 50191 79 53.11%
N/A Columns Drop 3,772  65 12.34%
New Columns Addition 3,772 69 12.25%

Energy Dataset:

Data Wrangling Step Rows Columns Missing Values
Raw Dataset 21812 130 67.32%
N/A Columns Drop 3795 66 36.68%
New Columns Addition 3795 31 0.18%

Data Merging Summary:

  • Number of Rows before merge: 3,772 (CO2 data) + 3795 (Energy Data)

  • Number of Records After Merge: 3,795

  • Key Matching Variable(s): country and year

Exploratory Data Analysis (EDA) Adjustments:

  • Normalization and scaling applied to numerical columns like: electricity_generation, renewables_electricity, fossil_electricity, gdp, and co2.

  • Created new columns in the energy dataset representing the share of electricity coming from different sources like: fossil_share_elec, solar_share, wind_share, and hydro_share.

  • Created new columns in the CO2 data set to show year over year change in CO2 emissions (co2_pct_change) and to show CO2 emissions from fossil fuels (fossil_fuel_co2)

linear models:

Regressing CO2 in the US vs every other column:

# 1. Filter for United States
us_data <- energy_co2_merged %>%
  filter(country == "United States") %>%
  select(-country, -iso_code, -continent, year)  # Remove non-numeric/grouping vars

# 2. Remove columns with all NA or zero variance
us_data <- us_data %>%
  select(where(is.numeric)) %>%
  select(where(~ sum(!is.na(.)) > 0)) %>%
  select(where(~ sd(., na.rm = TRUE) > 0))

# 3. Build models and extract p-values
results <- map_dfr(
  setdiff(names(us_data), "co2"),
  function(var) {
    df <- us_data %>% select(co2, !!sym(var)) %>% drop_na()
    if (nrow(df) < 10) return(NULL)  # skip if not enough data
    model <- lm(co2 ~ ., data = df)
    tidy(model) %>%
      filter(term != "(Intercept)") %>%
      mutate(variable = var)
  }
)

# 4. Print significant predictors (p >= 0.05)
significant <- results %>%
  filter(p.value < 0.05) %>%
  arrange(p.value)

head(significant %>% arrange(desc(estimate)))

Appendix F: Extra Visualizations

CO2 Regressed on every other column:

Appendix G: File Structure